How to use electronic configuration?

How can I use Git as a distributed, private repository for server configuration files?

• I am a new user of Git. I am thinking of controlling changes for configuration files in my Linux boxes, mostly CentOS. Putting all important configuration files for a box under one repository/project. Since all those configuration files are obviously scattered under different directories, how can I do that? Should I put root directory as the directory where I init Git? Should Git be the elegant solution, or is there a better solution? My idea is to push and pull configuration files from a server with ease and no need to manually backup or copy-and-paste to other location for backup. Don't even need to remember config file's location. Pull-edit-commit-push-done. Here I am looking for suggestion on the high-level and in detailed how-to's...

• Answer:

  What is your prime motivation for this? Is it to backup the files in case of a system failure or just keeping the old versions in case you need to revert one of them? If the first, this really isn't the correct way simply because it's highly likely you will forget something important so just reinstalling & fetching those files won't be enough. You still need to find your backup files. If you have plenty of files scattered around, something like http://www.yolinux.com/TUTORIALS/LinuxTutorialRCSintro.htmlmight work better. Another possibility is http://www.nongnu.org/rdiff-backup/. If you insist on using Git I would create a simple script that copies files that you want to control to a directory with correct path structure and then pushing that to a remote server. This way reverting is simple. If you just want to backup one or two directories (such as /etc) then Git would work better. The Git method, since you wanted hands-on examples, would look something like Create the "central" repository. Clone that repository to your local box. Write a script that cp's all the files with their path to that repository, cd to repo's directory, git commit && git push. And that's it. The problem is collecting all the necessary files in step 3. Of course this is just pushing, for pulling you need to do a simple cp (cd / && sudo cp -rf /my_backups/* .) after you pull the repo. You don't really want to create the repo in the root directory, that's why this is so complicated. Another problem is that this does not keep file owners and permissions. For that you need something like https://github.com/pangloss/gibak. Another that might work is https://gitorious.org/coldstorage.